Research On License Plate Recognition Algorithm In Natural Background

With China’s growing economic strength, a large number of private cars come into the lives of ordinary people. The management of motor vehicles has become an urgent problem for urban part. As traditional manual management is of costly and inefficient, the introduction of computer technology for automate management of traffic management has become a new trend. Before driving on a road motor vehicles usually have to be registered in the government department, legally registered vehicle will have a unique license plate number and vehicle license plates will be placed in a prominent position. For automated vehicle management system to achieve the primary task is to get a license number of the vehicle, usually such systems with hardware monitoring devices that can record images of vehicles on the road and through the analysis of these photographs which can detect illegal number plates.This thesis chose the license plate location, license plate character segmentation and character recognition as three key points were studied. The first is how to locate the license plate and split it from a monitor image, since there are differences in the natural color of the background noise difference in brightness and hardware devices in the whole image; besides, there are many parts of the image which have many similar characteristics to real license plate region, such as how to filter the ad text and other interference information region. The main concerns with segmentation problem are license plate and camera angles produce different and connected phenomena caused by frame material in the binary image, how to remove the interference and accurately divide the characters is an issue to be considered. The characters got before are actually character images, so it is necessary to translate them into digital type that computers can recognize. In order to achieve this goal, this paper refers to the research results of optical character recognition technology.Positioning license plate makes use of edge detection and mathematical morphology processing technology through a dynamic threshold and weighted gradient values successfully improves the positioning accuracy. And Radon transform is applied to correct the tilt and character stroke width factor is successfully used to eliminate borders connectivity issues. For the internal character of the connection, the ideal character width value is applied to segment. Character recognition makes use of support vector machine technology. Through the training of small sample data, it can identify the license plate characters.